When AI Builds Its Own Successors
Bernhard Liebl
5 min. read More than 80 percent of the code in Anthropic’s own development pipeline is now authored ...
Read Article
7 min read
29 percent of cloud spending is wasted. That’s over $100 billion globally each year – vanishing into idle resources, oversized instances, and forgotten subscriptions. The FinOps movement was meant to fix that. Yet the State of FinOps Report 2026 – three years after its mainstream breakthrough – reveals a sobering reality: only 14 percent of organizations have reached the maturity level where waste actually declines. For the remaining 86 percent, FinOps exists as a process – but delivers no results.
TL;DR
- 29 percent cloud waste: Nearly one-third of cloud spend delivers zero business value. This share has barely declined since 2023 (Flexera State of the Cloud 2026).
- Only 14 percent at “Run” maturity: Most organizations remain stuck at “Walk.” Without automation and governance, FinOps stays a reporting tool – not an operational discipline (State of FinOps 2026, n=1,192).
- 98 percent now manage AI spend: In 2024, just 31 percent did. This explosive jump signals that AI costs are surging faster than organizations’ ability to control them.
- 40 percent less waste for mature organizations: The difference between “Walk” and “Run” is measurable – and substantial (Flexera, 2026).
- 78 percent of FinOps teams report to CTO/CIO: Accountability is clear. The problem isn’t ownership – it’s execution.
29% – the number that hasn’t dropped in three years
The Flexera State of the Cloud Reports have documented the same uncomfortable truth since 2023: 27-29 percent of cloud spend is wasted. The figure refuses to budge. 2023: 27 percent. 2024: 27 percent. 2025: 28 percent. 2026: 29 percent. The trend, if anything, points slightly upward.
What is FinOps? FinOps (Financial Operations) is a management discipline that unites technology, finance, and business teams to make cloud spending transparent, reduce waste, and align investments with business outcomes. The FinOps Foundation defines three maturity levels: Crawl (initial visibility), Walk (active, manual optimization), and Run (automated governance). The goal isn’t minimal spend – it’s maximum business value per euro invested.
The primary drivers of waste have remained unchanged for years: idle compute (35 percent), oversized instances (25 percent), and unused commitment discounts. Companies buy Reserved Instances for workloads that never reach projected utilization. Developers provision test environments – and forget them after the sprint ends. And no one shuts down the staging instance that’s been running since October.
14% at the finish line – why the maturity curve is stalling
The State of FinOps 2026 Report (1,192 respondents, $83 billion in managed cloud spend) paints a sobering picture of FinOps maturity: Only 14.2 percent of organizations have reached the Run stage – the level where optimization is automated and embedded in governance processes. A full 51.4 percent remain stuck at Walk: they have visibility and perform manual optimizations – but lack automation.
Sources: State of FinOps 2026 (n=1,192), Flexera State of the Cloud 2026
The gap between Walk and Run is quantifiable: Organizations with mature FinOps practices report 40 percent less cloud waste than those at lower maturity levels (Flexera, 2026). Those 14 percent aren’t just saving money – they’re investing more precisely. So what’s holding back the other 86 percent?
The answer isn’t technical. The tools exist. Dashboards are built. The real barriers lie in three areas no dashboard can solve: organizational structure, incentives, and speed.
AI Spend: From 31% to 98% in Two Years
The most dramatic signal in the State of FinOps 2026 report is the explosion in AI cost accountability: In 2024, only 31 percent of FinOps teams managed AI spend. In 2025, it was 63 percent. In 2026, it’s 98 percent. In just two years, AI cost management has gone from niche topic to universal mandate.
The problem? Tools are lagging. The single most requested feature across the entire report is granular monitoring of AI spend – token consumption, LLM requests, GPU utilization per application. No commercial tool currently delivers this depth of insight. As a result, CIOs fly blind on AI spend – even as costs climb exponentially.
According to survey respondents, visibility into AI costs is the top challenge – followed by attribution to business units and ROI measurement. If you don’t know what a single AI use case costs, you can’t decide whether the investment pays off. And if you can’t decide that – you don’t optimize. You hope.
Three Reasons FinOps Fails to Scale
1. The incentive problem: Engineering teams are measured on feature velocity – not resource efficiency. Provisioning an instance one size smaller earns no recognition. Causing a production outage due to under-provisioned resources triggers an incident review. The rational decision at team level is over-provisioning – even though it generates 29 percent enterprise-wide waste.
McKinsey estimates the savings potential from rigorous FinOps implementation at 20-30 percent of cloud spend. For a company with a €10 million cloud budget, that’s €2-3 million annually. The math is simple. Yet execution still founders on the same organizational hurdles identified back in 2023. And with AI costs growing exponentially, the gap between potential and reality is widening – not narrowing.
2. The organizational problem: 78 percent of FinOps teams report to the CTO or CIO. That sounds like resolved accountability. But it isn’t. The teams generating the spend – engineering, data science, ML engineering – sit in separate reporting lines. FinOps can create transparency – but cannot cut budgets. Without binding consequences, dashboards remain decorative.
3. The complexity problem: Multi-cloud environments, AI workloads with variable token consumption, spot instances, committed-use discounts with differing terms – the hyperscalers’ pricing models are deliberately complex. The more services an organization uses, the harder optimization becomes. And every new AI service adds complexity faster than FinOps maturity grows.
The counterargument: FinOps does work – just not everywhere at once. The 14 percent of Run-level organizations prove it’s possible. They typically share one trait: FinOps metrics baked into engineering OKRs – and automated rightsizing policies that trigger without manual approval. The difference lies not in the tool – but in organizational anchoring.
What the 14% Do Differently
Automation – not reporting: Mature organizations automate rightsizing, shutdown policies, and commitment decisions. The FinOps team doesn’t produce reports for someone to read – it defines rules that act automatically. The distinction? Reporting informs. Automation executes.
Engineering integration: FinOps metrics are part of sprint reviews and engineering OKRs. Cost efficiency isn’t an afterthought – it’s a design criterion. That works only when the CIO embeds it in engineering culture – not as a cost-cutting measure, but as a professional standard.
AI spend as its own category: Mature teams treated AI spend as a distinct cost category from day one – with dedicated budgets, thresholds, and governance. If you wait until the invoice arrives to gain visibility, you’ve already lost control.
Conclusion
29 percent waste on $100 billion in global cloud spend. FinOps has existed for years – but only 14 percent have reached the maturity level where it truly delivers. The problem is well known. The tools are available. Teams are in place. What’s missing is organizational rigor: automated policies – not dashboards; engineering OKRs – not quarterly reports; and an honest answer to why the waste rate hasn’t dropped since 2023. The 40 percent lower waste among mature organizations proves it’s possible. But it’s possible only with deep organizational embedding – not another tool.
Frequently Asked Questions
Exactly what constitutes the 29% cloud waste?
Per the Flexera State of the Cloud 2026 report, enterprises waste an average of 29% of their IaaS and PaaS spend. Top contributors include idle compute (35% of waste), oversized instances (25%), and unused commitment discounts. Also included: forgotten test environments, orphaned storage volumes, and unused SaaS licenses.
What distinguishes Crawl, Walk, and Run in FinOps?
The FinOps Foundation’s three maturity levels: Crawl means initial visibility – the organization knows what it spends. Walk means active, manual optimization – teams identify savings and implement them themselves. Run means automated governance – policies trigger automatically, and cost efficiency is embedded in engineering processes. Per the State of FinOps 2026, 14.2% are at Run, 51.4% at Walk.
How do you manage AI costs with FinOps?
AI cost management requires granular monitoring at the token, request, and GPU level. Top challenges cited in the State of FinOps 2026: lack of visibility into actual per-use-case costs, difficulty attributing spend to business units, and unclear ROI. Mature organizations treat AI spend as its own cost category – with dedicated budgets, thresholds, and governance.
Is FinOps worthwhile for mid-sized companies?
Yes – starting at roughly €100,000 in annual cloud spend, FinOps becomes economically justified. With an average waste rate of 29%, that represents at least €29,000 in savings potential. An initial cloud cost assessment typically takes 2-4 weeks. For mid-sized firms, a single dedicated FinOps engineer often suffices – no full team required.
Who owns FinOps – the CIO or the CFO?
Per the State of FinOps 2026, 78% of FinOps teams report to the CTO or CIO – a 18-percentage-point increase since 2023. That makes sense: cloud optimization demands deep technical understanding of workloads. The CFO sets budget guardrails – but operational control rests with the technology team. What matters isn’t the reporting line – but whether the FinOps team has authority to enforce optimization actions.
Editor’s Reading Recommendations
More from the MBF Media Network
Header Image Source: Pexels / kaboompics (px:5899079)